Four types of RPS (i.e. types 1Gn, 2Gn, 2G and 3G), all of which contain Glc, L-Rha, Galp, Galf and GalNAc, occur on strains of Streptococcus sanguinis, S. gordonii and S. oralis that are primary colonizers of teeth. Based on studies of S. gordonii and S. oralis, the related structures of these polysaccharides depend on a relatively small number of genes for allelic glycosyl or glycosylphosphate transferases that exhibit high homology and synteny across species lines. In spite of such similarities, the large operons (i.e. rps gene clusters) that contain these genes are not identical. Thus, in S. oralis, the first three genes (i.e. rmlA, rmlC and rmlB) for TDP-L-Rha biosynthesis occur within the rps gene cluster whereas in S. gordonii, they occur in a separate operon that includes the gene galE2 for a bifunctional epimerase that converts UDP-Glc to UDP-Gal and UDP-GlcNAc to UDP-GalNAc. The last gene for TDP-L-Rha biosynthesis (i.e. rmlD) is transcribed monocistronically from a separate locus. In S. oralis, this gene is immediately downstream of the rps gene cluster whereas in S. gordonii, it is downstream of the rml-GalE2 operon. The molecular differences noted between S. oralis and S. gordonii raise the interesting possibility that each RPS-producing species has a characteristic arrangement of genes for RPS biosynthesis. To assess this possibility, we have identified the genes for RPS biosynthesis in S. sanguinis SK45, a previously unstudied species. We have also examined the location and arrangement of rml genes and galE2 in uncharacterized strains of S. gordonii that produce type 3G RPS, and strains of S. oralis that produce type 2Gn RPS. The rps cluster of S. sanguinis SK45 more closely resembled that of S. gordonii 38 than those of previously characterized S. oralis strains. As in S. gordonii, the rml genes of S. sanguinis were not in the rps cluster but instead occurred within separate operons, one that contained rmlA, rmlC, rmlB and the other that consisted of monocistronic rmlD. However, the gene galE2 was not found downstream of rmlB, as in S. gordonii, but instead, was at the 3-prime end of the S. sanguinis rps cluster, immediately downstream of wefE for the glycosyltransferase that completes synthesis of the type 1Gn repeating subunit. The arrangement of these genes in other RPS-producing strains of S. sanguinis followed the pattern seen in strain SK45. Thus, it may be possible to distinguish RPS-producing strains of S. sanguinis from those of other species by the location of galE2 with respect to rmlB and wefE. Further evidence that the arrangement of the rml genes and galE2 can be used to identify different RPS-producing species came from studies of type 3G RPS-producing S. gordonii SK120 and type 2Gn-RPS producing S. oralis SK92. The location and arrangement of these genes was S. gordonii-like in strain SK120 and S. oralis-like in strain SK92. Differences between species in the locations of specific genes for RPS production support the possible development of rapid PCR-based methods for the identification of these bacteria. The development of such methods would facilitate the characterization of the biofilm communities associated with different RPS-producing species, for example, type 1Gn RPS producing strains of S. sanguinis and S. oralis. These bacteria differ physiologically and thus, may well reside in distinct biofilm communities. Previous findings revealed a close relationship between the gene clusters for S. oralis type 4Gn RPS and S. pneumoniae CPS 10F. However, the predicted properties of certain common glycosyltransferases were not compatible with the available structure of CPS 10F, thereby raising the possibility that this structure is incorrect. To examine this possibility, we purified CPS 10F from S. pneumoniae 34355, the strain used in molecular studies, and determined the structure of this polysaccharide by high resolution NMR. Several errors in the previously determined structure were noted. The amended structure of CPS 10F was supported by results from carbohydrate engineering of type 4Gn RPS with selected genes from S. pneumoniae. The results of this work provide insight into the structural and molecular basis RPS-mediated recognition. Whereas type 4Gn RPS is recognized by type 2 fimbriated Actinomyces spp., CPS10F is not. The failure of CPS 10F to function as a coaggregation receptor can now be explained by the fact that the beta1-6 linked Galf moiety in the trisaccharide structure Galf-beta1-6GalNAc-beta1-3Gal is not in the linear polysaccharide chain, but instead occurs as a side branch. The formation of this branch depends on the action of the S. pneumoniae polymerase (Wzy), which links the reducing end of each CPS hexasaccharide repeating unit to the adjacent unit through subterminal GalNAc-beta of Galf-beta1-6GalNAc-beta1-3Gal rather than through terminal Galf-beta, as in S. oralis RPS. The critical contribution of internal beta1-6 linked Galf to receptor recognition was further established by engineering a linear surface polysaccharide in S. oralis that was identical to RPS except for the presence of a beta-1-3 linkage between Galf and GalNAc-beta1-3Gal. This engineered polysaccharide was not recognized as a coaggregation receptor, thereby indicating that a beta1-6 linkage from Galf is required for recognition of adjacent GalNAc-beta1-3Gal in wild type RPS. These findings illustrate a molecular approach akin to site-directed mutagenesis for relating bacterial polysaccharide structure to function. They also suggest the critical role of a single gene (i.e. wzy for the polymerase) in the possible evolution of S. pneumoniae CPS to S. oralis RPS.